Apply Checkpoint, LoRA, VAE

to apply checkpoint, lora, and vae , you must first download the relevant files while it is possible to generate textures using the stable diffusion web ui alone, these model files must be downloaded and applied for optimal quality the guide below details how to download and apply the models download the models civitai civitai is a stable diffusion ai model sharing platform you can search for multiple models, read reviews, and download them civitai https //civitai com/ hugging face hugging face https //huggingface co/ hugging face of stability ai github stability ai’s github apply model apply checkpoint to use the downloaded checkpoint file in the stable diffusion web ui and ai texturing in umodeler x, you need to place the file in a specific path and apply it copy the downloaded checkpoint file to the stable diffusion webui/models/stable diffusion folder alt text in the stable diffusion web ui, refresh the list by clicking the button next to the stable diffusion checkpoint area in the top left corner alt text after refreshing, you can see the newly added checkpoint in the drop down menu alt text apply lora the lora file must also be placed in a specific path to be applied copy the downloaded lora file to the stable diffusion webui/models/lora folder alt text in the stable diffusion web ui, under the generate button , click the show/hide extra networks button alt text click on the lora tab in the area shown below alt text click the refresh button on the right side to display the loras you put in the route alt text when you click on that lora, it will be entered into the prompt and the application will be complete alt text applying a vae you can also apply a vae file by placing it in a specific path, but the process is a little different copy the downloaded vae file to the stable diffusion webui/models/vae folder alt text in the stable diffusion web ui , click the settings tab alt text click on stable diffusion , then click on the refresh button in the sd vae alt text after refreshing, you will see the newly added vae in the drop down menu alt text after selecting the desired vae, click the apply settings button to apply it alt text ckpt and safetensors the models involved in stable diffusion are distributed in two formats ckpt and safetensors given the differences and issues with each format, we recommend downloading in the safetensors format whenever possible below we detail the differences between these two formats and the potential issues with the ckpt format ckpt format the ckpt format is used to store intermediate training states of a deep learning model (that's why it's called a checkpoint checkpoint ) this format makes it convenient to re experiment or tune the model later however, the ckpt format uses python's pickle pickle module to store data, which can have security issues info pickle module the pickle module helps python save data to a file in binary format ("serialization") and retrieve it back ("deserialization") it's a bit like canning food and taking it out later , and it ensures that you don't lose important data when you quit your program however, you need to be careful when deserializing data from untrusted sources due to security concerns with a maliciously written ckpt file , arbitrary code can be executed when opening (un pickling) the file , putting your pc at risk of attack below are some of the official documentation warnings for the pickle module warning the pickle module is insecure the pickle module is insecure unpickle only trusted data it is possible to construct malicious pickle data that executes arbitrary code when unpickled never unpickle data that may have come from an untrusted source or may have been tampered with safetensors format the safetensors format is a new file format that addresses security vulnerabilities in existing ckpt files while enabling faster model loading we recommend using the safetensors format and most newer models are now distributed in this format